Você está na página 1de 18

15398

J. Phys. Chem. 1996, 100, 15398-15415

The Electron Pair


Richard F. W. Bader,* S. Johnson, and T.-H. Tang
Department of Chemistry, McMaster UniVersity, Hamilton, Ontario L8S 4M1, Canada

P. L. A. Popelier
Department of Chemistry, UMIST, P.O. Box 88, SackVille Street, Manchester M60 1QD, UK
ReceiVed: May 7, 1996; In Final Form: July 5, 1996X

Eighty years have elapsed since Lewis introduced the concept of an electron pair into chemistry where it has
continued to play a dominant role to this day. The pairing of electrons is a consequence of the Pauli exclusion
principle and is the result of the localization of one electron of each spin to a given region of space. It is the
purpose of this paper to demonstrate that all manifestations of the spatial localization of an electron of a
given spin are a result of corresponding localizations of its Fermi hole. The density of the Fermi hole determines
how the charge of a given electron is spread out in the space occupied by a second same-spin electron,
thereby excluding an amount of same-spin density equivalent to one electronic charge. The Fermi hole is an
electrons doppelgangersit goes where the electron goes and vice versa: if the hole is localized, so is the
electron. The topologies of two fields have been shown to provide information about the spatial localization
of electronic charge: the negative of the Laplacian of the electron density, referred to here as L(r), and the
electron localization function ELF or (r). The measure provided by L(r) is empirical. It is based upon the
remarkable correspondence exhibited by its topology with the number and arrangement of the localized electron
domains assumed in the VSEPR model of molecular geometry. (r) is based upon the local behavior of the
same-spin probability, and it is shown that the picture of electron localization that its topology provides is a
consequence of a corresponding localization of the Fermi hole density. This paper provides a complete
determination and comparison of the topologies of L(r) and (r) for molecules covering a wide spectrum of
atomic interactions. The structures of the two fields are summarized and compared in terms of the characteristic
polyhedra that their critical points define for a central atom interacting with a set of ligands. In general, the
two fields are found to be homeomorphic in terms of the number and arrangement of electron localization
domains that they define. The complementary information provided by the similarities in and differences
between these two fields extends our understanding of the origin of electron pairing and its physical
consequences.

Relation between Electron Pairing and the Pair Density


Lewis1

demonstrated that the idea of considering a


In 1915,
pair of electrons as the basic unit in modeling the electronic
structure of a molecule provided a powerful tool for rationalizing
and understanding molecular geometry, bonding, and reactivity,
an idea that continues to play a dominant role in present day
chemistry. Lewis1 originally went so far as to postulate the
breakdown of Coulombs law to account for the pairing of
particles of like charge. We now understand the pairing of
electrons to be a consequence of the antisymmetrization
requirement imposed on the wave function by the Pauli
exclusion principle, as applied to electrons that possess two
measurable spin possibilities.2
The oft-quoted result of antisymmetrizing the wave function
with respect to the permutation of the space-spin coordinates
of every pair of electrons is that no two electrons with the same
spin can occupy the same point in space. In chemistry, however,
one is most interested in the spatial distribution of the electrons.
To determine the manner in which the exclusion principle affects
the electron distribution and its properties in real space, one
must determine how many pairs of electrons, on the average,
contribute to the electron density over the region of interest.
This information is given by the electron pair density, alternatively called the pair probability function.3
X

Abstract published in AdVance ACS Abstracts, September 1, 1996.

S0022-3654(96)01297-X CCC: $12.00

Pair Density. The electron density multiplied by an infinitesimal volume element, F(r) dr, gives the number of electrons
in dr and integrates to N, the number of electrons. Similarly,
the product of the pair density and a corresponding pair of
volume elements, F(r1,r2) dr1 dr2, gives the number of electron
pairs formed between these two elements, and its double
integration yields the number of distinct electron pairs, N(N1)/2. The density of R-spin electrons at r1 is FR(r1). The
uncorrelated pair density for simultaneously finding R-spin
density at r1 and r2 is given by the product FR(r1)FR(r2). The
correct correlated pair density everywhere is less than this, and
the Fermi hole determines the difference, a negative quantity,
between the correlated and uncorrelated pair densities for samespin electrons.3 This difference is a measure of the degree to
which density is excluded at r2 because of the spreading out of
the same-spin density originating from position r1, and thus the
Fermi hole density determines the manner in which density at
r1, in an amount equivalent to one electronic charge, contributes
to the pair density at other points in space. Pictorially, one can
imagine that as an electron moves through space it carries with
it a Fermi hole of ever changing shape, the density of the
electron being spread out in the manner described by its Fermi
hole and excluding density equivalent to one same-spin electron.4
Since the charge of the electron is spread out in a manner
described by the density of its Fermi hole, it follows, as
1996 American Chemical Society

The Electron Pair

J. Phys. Chem., Vol. 100, No. 38, 1996 15399

demonstrated by Bader and Stephens,4 that the quantum


mechanical requirement for the localization of an electron to
some particular spatial region in a many-electron system is that
the density of its Fermi hole be totally and maximally contained
within this region. In this limiting situation, all other samespin electrons are excluded from the region, and since the same
behavior is obtained for an electron of opposite spin in a closedshell system, the result is a spatially localized electron pair. The
Fermi hole density determines the local extent of exclusion of
same-spin density. An electron can go whenever its Fermi hole
is different from zero, and if the hole is localized, so is the
electron.
It is the purpose of the present paper to demonstrate that all
physical measures of electron localization or delocalization5 are
determined by the corresponding localization or delocalization
of the Fermi hole density. The Laplacian of the electron density
provides a physical, albeit empirical, measure of localization, a
consequence of its topology providing a mapping of the Lewis
pairs assumed in the VSEPR model of molecular geometry.6-8
ELF, the electron localization function of Becke and Edgecombe9 that is based upon a modeling of the same-spin
probability and thus the Fermi hole,10 is shown to provide an
equivalent description of electron localization. This correspondence is demonstrated by showing that the topological
structures that characterize these two scalar fields in molecular
systems are, in general, either homeomorphic or they are simply
derivable, one from the other.
Properties of the Density of the Fermi Hole. The principal
properties of the Fermi hole that underlie its role in determining
electron localization are easily understood in a pictorial manner,
and this is done here for the general reader, with the relevant
equations appearing in the Appendix and the details in the
literature.4,8 The discussion is for the density of electrons of a
given spin, say FR, as the Fermi correlation affects only samespin electrons. The density of the Fermi hole, like the pair
density itself, is a function of the coordinates of two electrons,
those of a reference electron, r1, and those of another, r2. The
Fermi hole density determines how the density of the reference
electron spreads out from r1 into the space of the second samespin electron. It is a negative quantity, since its value determines
the decrease in the same-spin density at each value of r2. The
density of the Fermi hole has two important properties: (a) its
magnitude equals the total density of same-spin electrons for
r1 ) r2. This ensures the complete removal of all other samespin density from the position r1. (b) Its integral over r2 equals
-1, corresponding to the removal of one same-spin electron
from the density, FR(r2).
At the Hartree-Fock level of theory, there is no correlation
between electrons of different spin, and the pair density is simply
the product of the densities of the R- and -spin densities
multiplied by a factor of one-half (so as not to count the same
pair twice), or

FR(r1,r2) ) (1/2)FR(r1)F(r2)

(1)

For same-spin electrons, however, the density at r2, FR(r2), is


reduced in value by the density of the Fermi hole, and the samespin pair density is given by

FRR(r1,r2) ) (1/2)FR(r1)R(r2)

(2)

where R(r2) is the difference between FR(r2), the spin density


at r2, and the magnitude of the density of the Fermi hole at r2
(eq A1 of the Appendix). Thus, R(r2) is a measure of the
amount of same-spin density not excluded from the position r2
by the spreading out of the density at r1 and is, therefore, the

Figure 1. Contour plot of the magnitude of the Fermi hole density


for LiH, together with its profile and that of the R-spin density along
the internuclear axis for three positions of the reference electron: at
the Li nucleus (a), at the bond critical point (b), and at the proton (c).
The R-spin density for LiH is also shown in (a) along with the boundary
defining the atomic basins of Li and H. The shaded area in a profile
map denotes the magnitude of the Fermi hole and the extent to which
other same-spin density is excluded at each value of the abscissa, |r1
- r2|. A difference exists between the spin density and the shaded
area over the entire axis in profile b, except for r2 ) r1. This difference
is the quantity R(r2) of the text. In (b) the hole and the density of the
reference electron are delocalized over the entire molecule. In (a) and
(c) the same quantities are localized within the basins of Li and H,
respectively, and R(r2) vanishes over the corresponding basin. The
contours of the Fermi hole density appearing in a neighboring basin in
(a) and (c) are too low in value to appear in the profile plots. The
contours are in atomic units, the outer one having the value 0.001 and
the remaining ones increasing inward in steps of 2 10n, 4 10n,
and 8 10n, with n beginning at -3 and increasing in steps of unity.

conditional probability of finding an R-electron at r2 when


another is known to be at r1.9 When R(r2) vanishes, the samespin pair density vanishes for the pair of coordinates r1 and r2,
and only the density of the reference electron contributes to
the R-spin density at r2.
A plot of the magnitude of the Fermi hole density for a fixed
location of the reference electron shows how the density of an
electron with coordinate r1 is spread out over the space of the
second electron, excluding an equivalent amount of charge, and
whether the Fermi density is localized or delocalized relative
to a chosen boundary. Such a plot is shown in Figure 1b for a
reference electron at the position of the bond critical point (bcp)
in LiH. It is to be compared with the plot of the R-spin density
FR(r) shown in panel a, a comparison made clearer by the profile

15400 J. Phys. Chem., Vol. 100, No. 38, 1996


of both quantities also shown in panel b. The difference
between these two plots is R(r2). The density of the Fermi
hole is delocalized over both atomic basins, as is the electron
itself. Aside from the point r1 ) r2, R(r2) is different from
zero and more than a single R-electron contributes to the pair
density over both atomic basins when the density of the
reference electron spreads out from the bcp.
This situation is to be contrasted with that shown in Figure
1a where the reference electron is placed at the position of the
lithium nucleus. The Fermi hole density in this case is localized
within the basin of the Li atom and over which, except for a
region of low density close to the bcp, it equals the total singlespin density. Since (r2) vanishes, all other same-spin electrons
are excluded from this region and the same-spin pair density
also vanishes. The hole is maximally localized in this case,
and its integral over the Li atom basin yields over 99% of the
charge of one electron. Since a -electron exhibits the same
behavior, a single R, pair of electrons determines the total
electron density within this basin when referenced to the Li
nucleus. Fermi correlation does not act directly to pair up
electrons. Rather, since there is no Fermi correlation between
electrons of opposite spin, an R, pair is obtained as a result of
all other electrons of both R- and -spin being excluded from
a given region of space.
What is perhaps remarkable, as is demonstrated in the
following, is that this same localized behavior and near total
exclusion of other same-spin electrons is obtained for almost
all positions of the reference electron within the basin of the Li
atom, with the result that 95% of the density within this basin
comes from a single R, pair of electrons, a localized electron
pair. The density of the Fermi hole is localized nearly equally
within the basin of the H atom (Figure 1c), and one has the
picture of a molecule wherein the density in real space arises
from the presence of a single pair of electrons in each of the
atomic basins, with only a small probability of exchange of
same-spin electrons between the two atomic basins, the exchange resulting from a delocalized hole as pictured in Figure
1b for reference electrons in or near the interatomic surface.
The goal is to find a method or methods to determine the
presence of such regions of spatial pairing within any molecule.
Measures of Spatial Localization of the Electrons. While
displays of the Fermi hole density are useful in determining its
basic localized or delocalized nature relative to FR(r),5,7,8,11,12
each display is for a single fixed position of the reference
electron. What is needed is a measure of the extent to which
some number of electrons are localized to a given spatial region.
Daudel and co-workers13-15 reasoned that there should be some
best decomposition of the real space of a system into a number
of mutually exclusive spaces called loges, with the best loges
yielding the most probable division of the system into localized
groups of electrons. It was proposed that the minimization of
the missing information function be used to define the best
loges,16 with the missing information function being defined in
terms of the quantum probabilities of some number of electrons
being found in a particular region of space.
In the first application of these ideas,17 it was found that the
same loge boundary that minimized the missing information
function also minimized the fluctuation in its average electron
population. This is an important finding, since the calculation
of the missing information function requires the full N-particle
density matrix while the particle fluctuation is determined by
just the pair density. Following on this work, Bader and
Stephens18 were able to show that the vanishing of the
fluctuation over a region of space requires that the Fermi hole
for each of its electrons be totally contained within its

Bader et al.
boundaries, work that leads directly to the demonstration that
the spatial localization of some number of electrons is determined by the corresponding localization of their Fermi hole
density.4
Establishment of the relation between the localization of the
electron and the density of its Fermi hole leads to a direct
physical measure of the extent to which some number of
electrons are localized to a given spatial region. One first
determines the total possible Fermi correlation contained within
a region , a quantity denoted by FRR(,), by integrating
the density of the Fermi hole weighted by FR(r1) over for all
possible positions of the reference electron within .4,8 This
procedure corresponds to the double integration of the exchange
density in Hartree-Fock theory. If NR() electrons are
completely localized within , then the Fermi correlation attains
its limiting value of -NR(), that is, a corresponding number
of same-spin electrons are excluded from . The ratio |F(,)/
N()| for electrons of either spin is thus the fraction of the
total possible Fermi correlation per particle contained within a
region , and this ratio, when multiplied by 100, is l(), the
percent localization of the electrons in .
The method of defining the contained Fermi correlation is
given an atomic basis by identifying the regions with the
basins of the topologically defined atoms in the theory of atoms
in molecules.8 The boundary enclosing an atomic basin, a
quantum mechanical proper open system,19a is defined as one
of local zero flux in the gradient vector field of the electron
density. Such proper open systems are identified with the
functional groups of chemistry,19b since they fulfill the observational requirement of exhibiting characteristic and transferable
sets of properties. The Fermi correlation over an atomic basin
and other atomic properties are calculated by using PROAIMV.20
All SCF calculations, including geometry optimization, are
performed by using GAUSSIAN 92 and the basis set 6-311++G(2d,2p) with six d functions.21 The Appendix lists the atomic
charges q() and percent localizations l(). Also listed are
the total energies, geometrical parameters, and properties of the
electron density at the bond critical point.
Extensive investigation of the contained Fermi correlation
shows that, aside from core regions and including ionic systems
such as LiH or BeH2, electrons are not, in general, spatially
localized to yield individual pairs4,17 (Table 6), the same systems
wherein it is possible to define loges that minimize the missing
information function.17 Unsurprisingly, the electron correlation
energy in such systems is found to be dominated by the intrapair
contributions.22 It is possible for some number of pairs of
electrons to be spatially localized within an atomic basin in ionic
systems. In NaF and MgF2, for example, the populations of
the atoms approach the limiting value of 10 for an LM core
and the localizations are in excess of 95% (Table 7).
Summary and Relation to the Orbital Model. The general
conclusion drawn from the spatial pairing exhibited by the pair
density in many-electron systems is that, aside from core regions,
the density is not characterized by the presence of strongly
localized pairs of electrons. This observation is not at variance
with the possibility of obtaining an equivalent description of a
system by a transformation to a set of localized molecular
orbitals, the pair density being invariant to the required unitary
transformation.4 Maximizing the contained Fermi correlation
for a region requires that each of the orbitals be localized to
a separate spatial region , so that the overlap of the orbitals
between two different regions vanishes.4,8 Thus, to achieve
localization of N electrons into N/2 distinct spatial pairs requires
that each orbital be separately and completely localized to one
of N/2 spatially distinct regions. While orbitals can appear

The Electron Pair

J. Phys. Chem., Vol. 100, No. 38, 1996 15401

Figure 2. Contour maps of L(r) and (r) for ClF5 (a) and SF6 (b) with (r) on the RHS. The contour values for L(r) in this and succeeding maps
begin at zero and increase (solid contours) and decrease (dashed contours) in steps of (2 10n, (4 10n, and (8 10n, with n beginning at -3
and increasing in steps of unity. The first dashed contour in (r) adjacent to a solid contour has the value 0.50 and decreases in steps of 0.05 away
from the solid contours. The first solid contour adjacent to a dashed contour has the value 0.55 and increases in steps of 0.05. The atomic
boundaries, as determined by the zero-flux surface condition, along with the bond paths are indicated in each diagram, with the bond cps being
denoted by dots. The values of the functions at the maxima and other cps evident in the contour maps are given in the corresponding tables. The
equatorial maxima in (r) for ClF5 are of limited extent and are nearly coincident with the corresponding bond cps. In L(r) they occur within the
basin of the Cl atom. For a third-row atom, the inner core concentration in L(r) is spikelike at the nucleus, while the innermost region of solid
contours in (r) is reduced to a point in the scale of these diagrams.

localized in their displays, they actually exhibit a considerable


degree of absolute overlap23 and they fail to satisfy the
requirement of separate localization. This must be the case in
the face of the properties determined for the pair density. The
ability to define a set of doubly occupied localized molecular
orbitals does not imply physical localization of the electrons
into spatial pairs. In ionic systems, even the canonical set of
orbitals is strongly localized to the separate atomic basins
yielding atomic sets of localized electrons. If the canonical set
is not localized, a unitary transformation to a localized set will
leave the properties of the pair density and its predicted lack of
electron pairing unchanged.
The Lewis Model As Displayed in the Laplacian of the
Electron Density
Clearly, electrons are not spatially paired to the extent perhaps
anticipated in view of the success of the Lewis model. The
topology of the electron density, while providing a faithful
mapping of the concepts of atoms, bonds, and structure,8,24 does
not give any indication of the bonded and nonbonded electron
pairs anticipated on the basis of the Lewis model. However,
as is well documented, the topology of the Laplacian of the
electron density, the quantity 2F(r), does provide a physical
basis for the Lewis model.6,7,8,25 The Laplacian of a scalar field
such as F(r) determines where the field is locally concentrated,
2F(r) < 0, and where it is locally depleted, 2F(r) > 0, with
the corresponding local minima and maxima providing features
that are absent from the topology of F(r) itself. Since electronic

charge is concentrated in regions where 2F(r) < 0, it is


convenient to define the function L(r) ) -2F(r), with a
maximum in L(r) then denoting a position at which the electron
density is maximally concentrated.
Empirically, one finds that the local maxima in L(r) provide
a remarkably faithful mapping of the localized electron domains
that are assumed to be present in the valence shell of the central
atom in the VSEPR model of molecular geometry.26 Not only
is there is a one-to-one correspondence in their number but also,
as importantly, in their angular orientation within the valence
shell of the central atom, as well as in their relative sizes. Thus,
one may make the assumption that the pattern of local charge
concentrations defined by the Laplacian of the electron density
denotes a corresponding pattern of partial condensation of the
electron pair density to yield regions of space dominated by
the presence of a single pair of electrons.
The Topology of L(r). The typical pattern of electron
localization revealed by the topology of the Laplacian distribution through the presence of bonded and nonbonded charge
concentrations (CCs) in the valence shell of a main group
element is illustrated in Figure 2a for the ClF5 molecule. One
notes that L(r) exhibits a shell structure for F and Cl wherein
each shell is characterized by a shell of charge concentration
(which appears as a spike at the position of the nucleus for the
inner shell) followed by one of charge depletion, behavior that
is characteristic of elements with Z < 40.8,27 For the remaining
elements, L(r) exhibits the number of maxima and minima
required by the appropriate number of shells, but the other

15402 J. Phys. Chem., Vol. 100, No. 38, 1996

Bader et al.

TABLE 1: Critical Points in L(r) and (r) for Molecules with Five and Six Electron Domainsa
molecule
ClF5 [6,8,4]

ClF3 [5,7,4]
SF6 [6,12,8]
SOF4 [5,6,3]L, [6,10,6]
SF4 [6,8,4]L, [5,9,6]

n(r,s)

L(r)

F(r)

(r)

F(r)

typeb

1(3,-3)
1(3,-3)
4(3,-3)
4(3,-1)
4(3,-1)
4(3,+1)
2(3,-3)
1(3,-3)
2(3,-3)
6(3,-3)
12(3,-1)
8(3,+1)
1(3,-3)
2(3,-3)
2(3,-3)
1(3,-3)
2(3,-3)
2(3,-3)

2.095
0.689
0.421
0.358
0.125
0.008
1.764
0.398
0.101
0.401
-0.137
-0.163
0.915
0.414
0.266
1.104
0.330
0.088

1.108
1.218
1.221
1.198
1.239
1.262
1.129
1.256
1.268
1.357
1.368
1.413
1.284
1.356
1.362
1.233
1.360
1.357

0.3944
0.2767
0.2400
0.2362
0.1924
0.1691
0.3646
0.2525
0.1964
0.2537
0.1316
0.1086
0.3999
0.2691
0.2254
0.2609
0.2586
0.1843

0.9882
0.8235
0.7209
0.6324
0.5238
0.4502
0.9860
0.7905

1.570
1.420
1.418
1.317
1.368
1.385
1.618
1.582

0.1926
0.2728
0.2246
0.2092
0.1656
0.1432
0.1718
0.2788

0.8467
0.3326
0.2703
0.8380
0.8438
0.8278
0.9961
0.8348
0.8178

1.595
1.453
1.482
1.483
1.574
1.804
1.801
1.651
2.093

0.3166
0.1214
0.0998
0.3706
0.3348
0.3718
0.1050
0.3546
0.2375

n
b to Fa
b to Fe
n-be
ba-be
n-be-be-ba
n
b to Fe
b to Fa
b to F
b-b
b-b-b
b to Oc
b to Fe
b to Fa
n
b to Fe
b to Fa

a n(r,s) denotes the number and type of critical point; r is radial distance from nucleus to cp. Values of L(r) and F(r) at the cp are in atomic
units, as is r. b n denotes a nonbonded maximum, b a bonded one. Subscript a (e) denotes an axial (equatorial) ligand. Maxima linked by (3,-1)
cp are joined by a-, as in b-b. Maxima linked in forming the face for a (3,+1) cp are indicated. c There are two such maxima in (r).

extrema do not undergo a sign change, with L(r) remaining


negative over the outer shell.28,29 The sign change reappears
for the noble gas elements at the close of each period. This
property of L(r) enables one to identify the shell of interest.
For example, in molecules containing an atom from the first or
second transition series, the ns electrons are transferred to the
ligands and the chemistry is determined by the distortions of
the outer shell of the core, which is clearly identified as the (n
- 1)th shell of charge concentration for which L(r) > 0.
The topology of a scalar field and its stability are succinctly
summarized in terms of its critical points,30 points where F(r) or L(r) ) 0, for example. The application of these ideas
to the topology of the electron density, coupled with the theorem
of structural stability of Palis and Smale31 and the catastrophe
theory of Thom,32 yields a theory of molecular structure and
structural stability.24 The same topological analysis can be
applied to the critical points in L(r) found in the outer shell of
charge concentration, generally the valence shell charge concentration (VSCC) of a main group element.8 A critical point
(cp) at position rc is labeled by the duo (r,s), where r, the rank,
is the number of non-zero curvatures of L(rc) and where s, the
signature, is the algebraic sum of their signs. The local maxima
or (3,-3) cps, whose presence denotes a bonded or nonbonded
charge concentration (CC), are linked one to another by the
unique pair of trajectories that originate at an intervening
(3,-1) cp, thereby generating the atomic graph, the analogue
of the molecular graph defined by the corresponding set of cps
in F(r). The atomic graph, in general, assumes the form of a
polyhedron bounding the nucleus in question. In general, each
critical point in L(r) for a given shell exhibits a negative radial
curvature, and its topology is consistent with the presence of a
(3,+3) or cage cp, a minimum in L(r), positioned at the nucleus
within the shell. The Poincare-Hopf relationship30 for the cps
defining an atomic graph is then readily transformed into the
more familiar polyhedral formula of Euler:

V-E+F)2

(3)

The vertices are defined by the (3,-3) cps or maxima in -2F


and the edges by the unique pairs of trajectories that originate
at (3,-1) cps and terminate at neighboring vertices, while the
(3,+1) or ring cps define the resulting faces of the polyhedron.
In some ionic or very polar systems the ring cps, at which
L(r) attains its lowest values, occur in regions where the electron

density is diffuse and small in value, with the result that certain
associated curvatures of L(r) approach zero and the cps are
topologically unstable. In these systems the ring cps can
bifurcate and be linked to further structure in a region of low
density that is of no physical significance. This instability is
not found for the (3,-3) or (3,-1) cps, and the form of the
polyhedron is, therefore, stable with respect to any bifurcation
occurring in its faces.
The atomic graph for Cl in the pseudooctahedral geometry
of ClF5 is illustrated in Figure 2a, and its critical points are
summarized in Table 1. It is an irregular octahedron, with the
positions of the six vertices corresponding to the presence of
one nonbonded and five bonded CCs. All of these cps are found
within the VSCC of Cl with characteristic radii of 1.1-1.2 au.
The geometry is dominated by the large nonbonded CC with
L(r) ) 2.10 au, whose presence causes the Fe-Cl-Fa bond
angle formed by the equatorial Fe with the unique axial Fa to
be less than 90 (Table 5). This feature is made particularly
clear in the envelope map of L(r) illustrated in Figure 3a. Also
in accord with the VSEPR model, the bonded CC facing Fa is
larger than those facing the equatorial fluorines (Table 1). Each
of the latter bonded CCs is linked to the bonded CC directed at
the axial F and to the nonbonded CC to yield a polyhedron
with eight edges and four faces (Figure 3), with the corresponding critical points satisfying eq 3 with 6 - 8 + 4 ) 2. The
number of each kind of critical point that satisfies Eulers
equation, the characteristic set,8 is denoted by [V,E,F] (Table
1). We shall find in the following that Cl exhibits the same
characteristic set [6,8,4] in ELF and the topologies of ELF and
L(r) are homeomorphic.
The properties of L(r) recover not only the geometrical
aspects of Lewiss model, but also his definition of an acidbase reaction. A nonbonded CC is a Lewis base serving as a
nucleophilic center, while a ring cp, since it defines the point
where electronic charge is least concentrated in its atomic graph,
is a Lewis acid serving as an electrophilic center.8 L(r) is
usually negative at a ring cp, and this corresponds to the
presence of a hole in the VSCC of the acid, that is, in a face
of the polyhedron representing its atomic graph. A Lewis acidbase reaction corresponds to aligning a CC on the base with a
hole on the acid, with the alignment of the two cps providing
a guide to their relative angle of approach. In an ionic crystal,
the vertices of the anion are directed at the faces of the cation,33

The Electron Pair

J. Phys. Chem., Vol. 100, No. 38, 1996 15403


The Lewis Model As Displayed in ELF
ELF as a Measure of Localization of the Fermi Hole. The
electron localization function (ELF) defined by Becke and
Edgecombe9 is based upon the properties of a function
introduced by Tal and Bader42 in a study of functionals for the
kinetic energy density. The function is a measure of the
difference between two kinetic energy densities:

(r) ) G(r) - g(r)

(5)

G(r) is the positive definite form of the kinetic energy density


that can be expressed as43a

G(r) ) (p2/8m)iFi(r)Fi(r)/Fi(r)

(6)

where Fi ) |i|2 is the density of orbital i (natural or HartreeFock) with occupation number i. The function g(r), the
Weiszacker functional,43b is expressed in terms of the total
electron density F(r):
Figure 3. Displays of L(r) and ELF or (r) in terms of isovalued
surfaces or envelope plots. (a) The surfaces of L(r) ) 0.38 au isolate
the six maxima in the VSCC of Cl in ClF5: the upper broad nonbonded
maximum, the axial bonded maximum, and the four smaller equatorial
bonded maxima surrounding the inner core of Cl. (b) The surfaces of
(r) ) 0.65 showing the inner core and the same set of maxima that
exhibit the same relative sizes as found in L(r). The bonded maxima
in this case are linked to the surfaces of similar value enveloping the
fluorine ligands. The polyhedron defined by the characteristic set
[6,8,4] is shown as an inset. Panels c and d are corresponding displays
of L(r) and (r) for SF6 for envelopes equaling 0.0 and 0.8, respectively.
The ligands appear in both displays, in addition to the six bonded
maxima and the inner core of S. The characteristic polyhedron in this
case is [6,12,8], as displayed in the inset. The surfaces in L(r)
encompass regions within which L(r) > 0 and within which the electron
density is concentrated.

and the same interactions account for the layered geometry of


solid chlorine.34 The activation of an oxygen atom in the surface
of MgO toward the adsorption of CO is the result of the
formation of a hole in its VSCC by the introduction of vacancies
generated by doping with cations of reduced charge35 or by
electronic excitation.33
Numerous studies have shown a correlation of the magnitude
of the nonbonded CC with experimental proton affinities.36-39
Base strengths and preferred sites of protonation have been
similarly rationalized.40,41 The correlation has been shown to
break down in cases where charge dispersal in the products
dominates the reaction, as found in the protonation of fluoroand chlorophosphines.39 There is an energetic basis underlying
these correlations, a result of L(r) appearing in the local
expression for the virial theorem:

(p2/4m)2F ) 2G(r) + (r)

(4)

where G(r) > 0 is an electronic kinetic energy density defined


in the following section and (r) < 0 is the electronic potential
energy density defined by the virial of the forces exerted on
the electrons. A local CC, a Lewis base, is thus a region of
space where the local energy is stabilized by a dominant
potential energy contribution. A Lewis acid-base reaction is
the combination of a region with excess kinetic energy with
one of excess potential energy, with the excesses being measured
relative to the values required for the local satisfaction of the
virial relation obtained where L(r) ) 0.

g(r) ) (p2/8m)F(r)F(r)/F(r)

(7)

It is clear from eq 6 that g(r) is the correct form for the kinetic
energy density only for a single electron or a two-electron
Hartree-Fock ground state, that is, for a system described by
a single orbital. The difference between these two functions
for any other system vanishes at a nuclear position, r ) 0, and
for positions far removed from nuclei, r w . Tal and Bader42
demonstrated two other important properties of : (a) that (r)
g 0 and thus g(r) underestimates the kinetic energy density for
0 e r e and (b) that the requirement for (r) to approach
zero is for each orbital and, thus, each orbital density Fi to be
localized to its own spatial region. They noted that this is the
very condition required for the spatial localization of the Fermi
hole density and for the creation of electron pairs.4 Thus, one
has the important result that the kinetic energy density contains
information regarding the spatial localization of the electrons.42
Consequently, they found g(r) to be a good model of the kinetic
energy density in LiH,42 where the orbitals are spatially localized
and where, as discussed with reference to Figure 1, electron
pairing is very pronounced in each atomic basin with l(Li) )
l(H) ) 95% (Table 6).
The function (r) can be defined for each set of same-spin
electrons, with R(r) ) (r)/2 for a closed-shell system. In a
region of space over which R(r) approaches zero, the Fermi
hole density thus is correspondingly localized, approaching -FR(r) in value, and a single electron of given spin will determine
the kinetic energy density. If each orbital in a many-electron
system is so localized to a distinct spatial region, then the
expression for G(r) becomes a sum over separate spatial regions.
There will be one region for each Fi(r) over which it will equal
the density F(r) appearing in the expression for g(r) and hence,
in this limit, G(r) will equal g(r).
This situation is to be contrasted with the ground state of a
system of non-interating Bosons in which all of the particles
occupy the same orbital and every particle is equally delocalized
over the entire system. A Bose-Einstein condensate has
recently been observed in the cooling of a cloud of integerspin 87Rb atoms that closely approach the non-interacting
condition.44 When the de Broglie wavelength becomes larger
than the mean spacing between the particles, the wave functions
for the atoms overlap to such an extent that individual atoms
can no longer be distinguished and Bose statistics favors the
condensation of all of the particles into a single quantum state,
which is just the opposite extreme to the situation required for

15404 J. Phys. Chem., Vol. 100, No. 38, 1996

Bader et al.

the spatial pairing of electrons. Thus, g(r) describes the kinetic


energy density in both limiting situations: for Fermions, F(r)
is given by a sum of spatially separated orbital densities, one
for each electron pair, while for Bosons, F(r) is given by the
density of a single orbital extending over the entire system and
occupied by all of the particles.
In a modeling of the exchange energy, Becke10 showed that
the leading term in the Taylor expansion of the spherically
averaged conditional pair probability, the quantity R in eq 2,
is proportional to R(r). Beckes finding is understandable, for
the vanishing of (r) corresponds to the Fermi hole density
attaining its limiting form of -FR(r2), the same requirement for
the vanishing of R(r2). What we emphasize here is that the
degree of electron exclusion determined by the smallness of
R(r) is a consequence of the spatial localization of the density
of the Fermi hole.
Becke and Edgecombe9 note that R(r) itself is not suited
for a direct determination of the degree of electron localization;
a low value of R(r) denotes a high localizability and R(r) is
not bounded from above. They, therefore, propose the function
called ELF, denoted here by (r), where

(r) ) (1 + 2)-1

(8)

where ) R/Ro and Ro is the quantity corresponding to R


but for a uniform electron gas with the same spin density as
the system in question. The range of values for (r) is thus
restricted to 0 e (r) e 1, with (r) ) 1 denoting complete
localization and (r) ) 1/2 corresponding to uniform electron
gaslike pair probability. It is important to note the more-orless arbitrary9 nature of the transformation used to obtain a
function with these bounded values, a feature that will require
comment in the comparison of the topologies of L(r) and (r).
Becke and Edgecombe9 find that (r) defines atomic shell
structure up to and including the shell for n ) 6 and offers
remarkably clear three-dimensional realizations of core, bonded,
and nonbonded regions in molecular systems. They note that
the Laplacian of the electron density provides alternative
definitions of the same features, and this raises the question as
to whether the two functions L(r) and (r) complement one
another or provide alternative but equivalent descriptions of the
spatial localization of the electron density.
Savin and co-workers have made extensive applications of
ELF for solids with the diamond structure45a and for a wide
variety of molecules45b using computer color graphics to display
the results. More recently, Silvi and Savin46 have demonstrated
how the local maxima in (r) may be isolated through a study
of the topological behavior of isovalued surfaces47 of (r). This
topological approach to identifying the (3,-3) cps or attractors
in a scalar field is most easily understood in terms of an
example. Consider two attractors in F(r) situated at the nuclear
positions of a diatomic molecule and linked, as they necessarily
are, by a (3,-1) or bond cp where the value of the density is
denoted by Fb. This is a saddle point in a two-dimensional
display of F(r) in a plane containing the internuclear axis. A
single isovalued surface in the electron density will be obtained
for all values of the density F(r) < Fb, with the envelope
retaining its generic shape as the value of the density in the
surface approaches the value Fb but becoming increasingly
pinched in the vicinity of this critical point. The isovalued
surface with density equal to Fb is, however, unstable, with any
further increase in F(r) causing it to bifurcate into two surfaces,
each enclosing a single attractor. This pattern of two isolated
attractor domains persists for all further increases in the density.
Silvi and Savin characterize the topology of the isovalued
surfaces of ELF by defining localization domains:46 a surface

in (r) containing more than one attractor is called a reducible


domain, whereas a domain containing only one attractor is called
irreducible. Thus, as the value of (r) in the surface defining
a reducible domain is increased from some initial value it
undergoes bifurcations at critical values of (r) (the values of
its (3,-1) cps) into two or more domains containing fewer
attractors. This process may be continued until all domains
attain their irreducible form. The irreducible localization
domains, since they isolate the local maxima in ELF, correspond
to core, bonded, and nonbonded attractors, as originally identified by Becke and Edgecombe.9 Still more recently, Fassler,
Haussermann, and Nesper48 use displays of ELF to visualize
free surface bonds and bonds connecting surface atoms for
various models for the Si(100) surface.
Topology of ELF. This paper presents a full topological
analysis of ELF, one that determines all of its critical points
and compares the structure so defined with the corresponding
structure exhibited by the Laplacian of the electron density.
Critical points other than the (3,-3) or local attractors of the
associated gradient vector field are of importance not only in
determining the structure of a scalar field but also in relating
the field to chemical properties, as demonstrated earlier for L(r).
Location of the cps of rank 3 in (r) and their signatures is
accomplished by using a newly developed version of the
program MORPHY 96.49 This is a new semiautomated program
for the application of the theory of atoms in molecules that uses
a superior technique to locate cps.50
The ELF distribution for the ClF5 molecule is illustrated in
the form of a contour map of (r) in Figure 2a, and its electron
domains are displayed in an isovalued surface map in Figure
3a. Its critical points are listed in Table 1. In all contour maps,
values of (r) > 0.5 are denoted by a solid contour and those
with (r) e 0.5 by a dashed contour, with the solid contours
denoting a localization greater than that for a uniform electron
gas with an identical spin density. The chlorine atom exhibits
an atomic graph consisting of the set [6,8,4] in both fields, and
they are homeomorphic for this atom. The similarity extends
to the relative values of the maxima, with the values of the
(3,-3) cps in (r) and L(r) decreasing in the order nonbonded
n, axial bonded b to Fa, and equatorial bonded b to Fe, in
agreement with the VSEPR model. Note the small size of the
equatorial bonded compared to the axial bonded domain in (r). The VSEPR model relates the displacement of the
equatorial ligands toward Fa to yield an acute value for the bond
angle Fe-Cl-Fa (Table 5) to the dominant size of the
nonbonded domain, an effect that is accounted for in the
interdomain angle n-Cl-be being obtuse in both fields,
equaling 93.2 and 98.3 in L(r) and (r), respectively, as
determined by the coordinates of the cps.
A comparison of Figure 2a,b illustrates the general features
that the two fields have in common, as well as their principal
differences. Both fields exhibit the same shell structure for both
atoms, three for Cl and two for F. The nonbonded domains
are, in general, of greater spatial extent in (r) than they are in
L(r), and the corresponding critical points are located farther
from the nucleus, 1.11 au in L compared to 1.57 au for . The
bonded domains are of more equal size. The same disparity in
size of the nonbonded domains is found for the ligands. Both
fields exhibit two nonbonded domains on each Fe, while Fa, as
a consequence of symmetry, exhibits four, those in (r) having
the greatest extent and being farthest removed from the nuclei.
Each Fe also exhibits a bonded domain in (r), but not in L(r).
The spread in the radial distances of the critical points for a
given atomic graph is, in general, smaller for L(r), where the

The Electron Pair

J. Phys. Chem., Vol. 100, No. 38, 1996 15405

TABLE 2: (3,-3) Critical Points in L(r) and (r) for Hydridesa


molecule
borane [3,3,2]
methane [4,6,4]
ammonia [4,6,4]
water [4,5,3]
hydrogen fluoride
silane [4,6,4]
phosphine [4,6,4]
hydrogen sulfide [4,5,3]
hydrogen chloride

L(r)

F(r)

(r)

F(r)

typeb

24.980
1.268
2.632
2.194
5.349
3.383
4.703
9.343
22.978
0.340
22.900
0.602
0.657
0.828
0.858

2.243
1.012
0.748
0.831
0.649
0.717
0.635
0.576
2.781
1.443
2.649
1.297
1.500
1.281
1.185

0.4178
0.2991
0.5562
0.4805
0.9427
0.7445
1.1169
1.4736
0.3879
0.1319
0.3865
0.1973
0.2211
0.2807
0.2730

>0.9500
>0.9900
0.9567
0.9994
0.9213
0.9981
0.9960
0.8896
1.0000
0.9954
1.0000
0.9694
0.9998
0.9994
0.9245

2.240
2.083
1.388
1.921
1.109
1.810
1.730
0.955
2.837
2.327
2.685
1.867
2.531
2.412
1.625

0.4170
0.4162
0.1584
0.4170
0.3242
0.4074
0.3856
0.5292
0.3694
0.0420
0.3758
0.0900
0.3692
0.3532
0.1404

b
b
n
b3
n2
b2
b
n (ring)
b4
n
b3
n2
b2
b
n (ring)

a See footnote a to Table 1. b A > sign before (r) means that the maximum is flat to the extent that its location could not be determined
precisely, and the values of F(r) and r refer to the position of the proton, its approximate location in other hydrides.

cps occur in a given shell of charge concentration with its own


characteristic radius, than it is for (r).
The L and atomic graphs for the sulfur atom in SF6 are
again homeomorphic with six domains, this time all bonded to
yield the characteristic set [6,12,8]. The same general comparative observations apply (Table 1 and Figures 2 and 3). In this
molecule the bonded domains, which in L possess a radial
distance characteristic of the VSCC of the free sulfur atom, fall
within the atomic boundaries of the F atoms in both fields, a
reflection of the more polar nature of the S-F interaction, with
q(F) ) -0.7e, as opposed to Cl-F, with q(F) ) -0.5e (Table
5). In general, we shall be concerned with the atomic graph
associated with the central atom only, one that can include
maxima lying within the atomic basins of the ligands as found
in SF6. Only data for the maxima in the two fields are tabulated
for the remaining molecules.
Comparative Morphology of L(r) and (r)
Second- and Third-Row Hydrides AHn. The L(r) and (r) fields for atom A are homeomorphic for all members of the
second- and third-row hydrides AHn (Table 2 and Figures 4
and 5). The general forms of the two fields for members of
both series with q(H) < -0.7e, Li w B and Na w Si (Table 6),
are represented by BH3. Note, as illustrated for BH3 and as
also found up to PH3 in the third row, that the interatomic
surfaces for these molecules fall on the natural boundaries
defined by the L(r) and (r) fields, a reflection of the high
degree of localization of the electrons within the basins of the
A atoms in these cases (Table 6). In all cases of an A atom
bonded to a hydrogen, the bonded domains in (r) are localized
on the the protons, behavior that is found for L(r) only in those
molecules with significant hydridic character, Li w B and Na
w P. In the remaining members, L(r) exhibits bonded maxima
within the basin of the A atom in addition to the maxima at the
positions of the proton that are necessarily present in all
molecules. The hydridic members illustrating bonded maxima
only on the protons in L(r) are those whose l(H) values are in
the range 70-95%, that is, in those cases where the hydridic
nature is reflected in a corresponding localization of an electron
pair within the basin of the H atom. The bonded domains in
(r) are also significantly larger than those in L(r). In LiH the
0.95 contour almost envelops the whole of the basin of the H
atom, closely mimicking the interatomic surface, a reflection
of the localization of the Fermi hole illustrated in Figure 1. The
same behavior is found for the other ionic members of both
series, and the association of high l(H) values for atomic basins

with (r) > 0.95 demonstrates that electron localization as


determined by ELF is a consequence of the localization of the
Fermi hole.
In methane, ammonia, silane, and phosphine, the four maxima
define a tetrahedral atomic graph with six edges and four faces.
For both water and hydrogen sulfide, however, the same number
of maxima now defines a polyhedron that possesses one less
edge, the one linking the bonded domains, and hence one less
face in both fields. This same atomic graph is found for the
oxygen atom in an ether and in the OH fragment of a carboxylic
acid. Sulfur exhibits the same atomic graph in thioethers and,
as discussed in the following, in SF2. Aside from maxima
associated with the protons in ionic systems, the critical points
in L(r), unlike those in (r), exhibit values and radii characteristic of the atom and the quantum shell in which they occur,
with the values of the maxima increasing and their radii
decreasing across a row of the periodic table. The values of
(r), on the other hand, provide a measure of the absolute degree
of localization associated with each attractor domain.
The VSEPR model assigns a larger size to nonbonded than
to bonded domains, as is generally observed in the L(r) field.
In H2O and H2S, the angle formed with the A nucleus by the
two nonbonded charge concentrations in L(r) exceeds the
tetrahedral angle, equaling 128 and 121, respectively, while
that between the two bonded charge concentrations is less than
this value, equaling 106 and 95, respectively, all as anticipated
on the basis of the VSEPR model. In (r) the same pairs of
angles are 109 and 121 for the nonbonded domains and 106
and 96 for the bonded domains.
The two fields are again homeomorphic for the A atom in
HF and HCl, where one encounters nonisolated critical points.24
An off-axis critical point in a linear molecule must necessarily
belong to a set of nonisolated critical points, i.e., a connected
set, which encircles the axis, each member of the set exhibiting
a single zero curvature tangent to the ring. Such critical points
are thus of rank 2. The radial curvature directed toward the
axis is negative, while the remaining one, the one parallel to
the axis, may be negative to yield a ring attractor, a connected
set of (2,-2) cps, or it may be positive to yield a connected set
of (2,0) cps. A (2,0) cp is equivalent to an edge or (3,-1) cp,
as it links neighboring attractors. Both HF and HCl possess an
axial bonded attractor and a nonbonded ring attractor with an
intervening set of (2,0) cps. Gradient paths of the corresponding
vector fields originate at both the axial and ring attractors and
terminate at the intervening ring of (2,0) cps. The remaining
trajectories originating at the ring attractor terminate at an axial
(3,+1) cp on the nonbonded side of the nucleus. The nonbonded

15406 J. Phys. Chem., Vol. 100, No. 38, 1996

Bader et al.

Figure 5. Contour plots of L(r) on the left and (r) on the right for
third-row hydrides: (a) SiH4, (b) PH3, (c) SH2, and (d) ClH. The plots
for SH2 are for the plane containing the two nonbonded domains. The
bonded domains are localized on the protons in all cases except for
L(r) in SH2 and ClH.

Figure 4. Contour plots of L(r) on the left and (r) on the right for
second-row hydrides: (a) BH3, (b) CH4, (c) NH3, (d) OH2, and (e) FH.
In (r) the bonded domains are localized on the protons, while in L(r)
this is true only in the ionic and polar cases such as BH3. The upper
axial cp in water is the (3,-1) cp linking the two nonbonded domains
in the perpendicular plane, a plane illustrated in Figure 5 for SH2.
Electron domains in both fields contract in size with increasing nuclear
charge on A.

domains in both fields are necessarily ringlike for linear


molecules and only slightly depart from this behavior in
situations with high axial symmetry. The pronounced nature
of the nonbonded ring attractor on Cl is typical and found in
Cl2 as well. Its presence imparts a very anisotropic shape to
the nonbonded axial hole or (3,+1) cp. It is the interaction of
the nonbonded attractor on one Cl with the hole on another
that accounts for the dominant intermolecular interaction found
in the layered structure of solid chlorine.34

Second- and Third-Row Fluorides AFn. The behavior of


the two fields for fluorides AFn is illustrated by the examples
given in Table 3 and Figures 6 and 7. The BF3 molecule
provides an example of a connectivity not observed in F(r),
wherein a pair of attractors is linked by two (3,-1) cps. The
three maxima in both fields for this molecule are so linked by
three pairs of (3,-1) cps, these being placed symmetrically on
each side of the symmetry plane to generate three faces. Both
fields possess the same polyhedron with three vertices and six
edges, but the set [3,6,6] does not satisfy Eulers formula
because of connectivity to outer structure.
The atomic graphs for C and N in their fluorides are regular
tetrahedrons for both fields, as found in their hydrides. The
bonded maxima in (r) for the molecules NF3, CF4, and BF3
appear shared between A and F, while for L(r) they appear
separate from the valence shell on F, becoming increasingly
smaller (N w B) and being entirely absent from the ionic
members MgF2 and LiF. In the less electronegative third-row

The Electron Pair

J. Phys. Chem., Vol. 100, No. 38, 1996 15407

TABLE 3: (3,-3) Critical Points in L(r) and (r) for Fluorides and PCl3a
molecule
BF3 [3,6,6]
CF4 [4,6,4]
NF3 [4,6,4]
PCl3 [4,6,4]
OF2 [2,3,3]L, [4,5,3]
SF2 [4,6,4]L, [4,5,3]
ClF2+ [4,5,3]
F2
a

L(r)

F(r)

(r)

F(r)

type

0.182
1.078
5.359
1.561
0.417
0.182
10.017

1.141
0.979
0.691
0.842
1.411
1.557
0.607

0.3155
0.3382
0.7515
0.4034
0.1476
0.1274
1.2195

0.8621
0.8824
0.9935
0.8515
0.9969
0.8643
0.9718
0.7732

1.330
1.154
1.333
1.126
2.076
1.979
1.064
1.157

0.4366
0.3824
0.2072
0.3774
0.0634
0.3422
0.4196
0.3710

1.006
0.088
1.724
0.496
-0.154
12.540

1.253
1.396
1.130
1.245
1.256
0.560

0.2489
0.2123
0.3594
0.2768
0.3669
1.6348

0.8178
0.9873
0.7922
0.7101
0.9207

1.864
1.639
1.496
1.256
0.941

0.3946
0.1604
0.2948
0.3669
0.5826

b3
b4
n
b3
n
b3
n2
b2
n2
b2
n2
b2
b
b (ring)

See footnote a to Table 1.

Figure 6. Contour plots of L(r) and (r) for second-row fluorides:


(a) BF3, (b) CF4, and (c) NF3 and for PCl3 (d).

Figure 7. Contour plots for L(r) and (r) for second- and third-row
fluorides: (a) OF2, (b) SF2, (c) ClF2+, and (d) F2.

atoms, no bonded CCs appear until SF2 (Figure 7). In PF3,


(r) has the set [4,6,4] characteristic of a tetrahedron as found
for NF3. L(r) exhibits the same tetrahedral structure, but while
the nonbonded CC is quite marked, the bonded maxima now
occur in regions where the density is depleted rather than
concentrated, and, in addition, no bonded CCs are found on
the fluorines. SiF4 exhibits similar behavior, with (r) having

the set [4,6,4], but in L(r) the four maxima occur in regions
where L(r) < 0. Thus, in very polar fluorides as well as in the
corresponding hydrides, only nonbonded CCs appear within the
VSCC the A atom in L(r).
The situation is different for the less polar third-row chlorides,
as evidenced by PCl3 where the net charge on P is decreased to
+1.8e from +2.5e in the fluoride (Table 7). Both fields exhibit

15408 J. Phys. Chem., Vol. 100, No. 38, 1996

Bader et al.

TABLE 4: (3,-3) Critical Points in L(r) and (r) for Single, Double, and Triple Bondsa
molecule
C2H6 [4,6,4]
C2H4 [3,5,4]L, [4,6,4]
C2H2
N2
CO O
C
H2CO
[3,3,2] O
[3,5,4] C
dimethyl ether
[4,5,3] O
[4,6,4] C
SO2 [3,4,3]
SO3 [3,6,3]L, [6,9,5]

L(r)

F(r)

1.034
1.282
1.478
1.405
1.560

0.976
1.010
0.977
0.994
1.115

0.2795
0.3004
0.3769
0.3122
0.4407

1.515
3.963
2.922
5.483
4.901
1.310
5.878
3.794
1.537
1.452
2.706
5.688
0.837
1.457
1.389
0.866
0.743
0.811

0.982
0.824
0.733
0.688
0.649
0.863
0.644
0.710
0.970
0.955
0.717
0.646
1.023
0.987
0.992
1.257
1.322
1.299

0.3255
0.7738
0.5779
0.9811
0.9165
0.3174
0.9741
0.8259
0.3198
0.4826
0.7034
0.9700
0.2805
0.3170
0.3095
0.2283
0.3866
0.3902

F(r)

type

0.9685
0.9999
0.9412
0.9999

1.440
2.088
1.332
2.063

0.2496
0.4206
0.3044
0.4224

0.8906
0.9999
0.8854
0.9811
0.8786
0.9379
0.9980
0.9320
0.8801
>0.9500
0.8801
0.9069
0.9256
0.9069
0.9500
0.9500
0.9939
0.8529
0.8348

1.443
2.015
1.008
1.342
1.002
1.148
1.645
1.103
1.104
2.063
1.123
1.331
1.116
1.299
2.040
2.040
1.819
1.532
1.506

0.2250
0.4072
0.7648
0.1656
0.6914
0.2760
0.0834
0.3340
0.5424
0.4387
0.5424
0.3198
0.3270
0.3198
0.4000
0.4000
0.1022
0.4630
0.3772

b to Cb
b to H
b to Cc
b to H
b to Cd
b (ring)
b to H
b1 in
n
b to C
n
n
n
b
b to H
b to O
b to C
n
b to O
b to H
b to H
n
b2
b3 in L
b6 in

(r)

a See footnote a to Table 1. b A central maximum in . c Two central maxima in . d Single central maximum in L. Ring attractor in is
centered at midpoint.

a [4,6,4] tetrahedral structure for P with the bonded domains in


L(r) appearing as distinct CCs, as found for the A nuclei in
CF4 and NF3 (Table 3).
The atomic graph for oxygen in OF2 for (r) is isomorphic
with that for both fields in H2O, but while the L(r) field appears
similar (Figure 7), what appear to be the two bonded charge
concentrations are actually (3,-1) cps. This behavior for singly
bonded oxygen is found for ligands more electronegative than
C in the second row and for S and Cl in the third. In SF2 there
are four attractors in both fields, and that for (r) is homeomorphic with that found for S in SH2 with three faces while
that for L(r) is a regular tetrahedron. The two fields are
homeomorphic for the molecule ClF2+, and the pattern is as
found for the oxygen and sulfur nuclei in the hydrides and for
sulfur in SF2.
The two fields exhibit the same basic structure in F2; both
possess a central bonded maximum and a ring of nonbonded
attractors (Figure 7). In (r) the bonded and nonbonded
attractors are linked by a (2,0) ring and in L(r) by a corresponding axial (3,-1) cp. Both fields exhibit an axial (3,+1)
cp, which serves as the terminus for the trajectories from the
nonbonded ring of attractors. The central maximum in F2 is
atypical. It is found in a region where L(r) is negative and it
is not a CC. However, a very large basis set does yield a
negative Laplacian at the bcp in F2.34
The l(F) values in Table 7 show that, regardless of its net
charge, the electron density on a fluorine atom not only is tightly
bound but is physically localized within its own basin. While
this leads to stability in ionic fluorides, the lack of exchange
with the electrons in a neighboring atomic basin results in weak
shared interactions, as found in F2. The separate accumulation
of density in each of the atomic basins in F2 is reflected in 2Fb
being close to zero (it is slightly negative with the use of a
very large basis set), with the final distribution being dominated
by the positive curvature and resulting stress parallel to the bond
path. While the Hartree-Fock description of F2 does not
account for its binding, the properties of the electron density
are relatively insensitive to any added correlation.

Single, Double, and Triple Bonds. The ethane, ethene, and


ethyne molecules provide a comparison of the two fields for
single, double, and triple bonds (Table 4, Figure 8). The bonded
maxima associated with the protons in L(r) in these molecules
are, like those in the covalent and polar hydrides, present as
separate charge concentrations in the VSCC of the carbon atom,
while in (r) the corresponding bonded domains are localized
on the protons. A carbon atom of ethane has the characteristic
set of a regular tetrahedron [4,6,4] in both fields, with the
difference that in L(r) each carbon has four bonded maxima
while in (r), in addition to the three domains associated with
the protons, only a single bonded domain midway between the
carbons is shared by both atoms (Figure 8).
The [4,6,4] set is characteristic of a saturated carbon for both
L(r) and (r) being found in methane, ethane, and also dimethyl
ether (Table 4). In the latter molecule, one again finds separate
bonded maxima on the carbon and the oxygen in L(r), while in
(r), a single maximum is shared between them. With this
understanding, then, the two fields are also homeomorphic for
the oxygen atom, which has the same characteristic set found
in H2O.
In ethene each carbon exhibits a bonded domain within its
own VSCC for the C-C link in L(r), while in (r) the bonded
domains, two in number, are located midway between the carbon
atoms, one on each side of the molecular plane. Thus, (r)
indicates the presence of a carbon-carbon double bond by
defining a corresponding number of bonded domains,46 but this
is not a general result for double bonds between other atoms.
This behavior is not observed in L(r), which, like F(r), indicates
the presence of a double bond through an increase in the value
of L(r) or F(r) at the critical point and by exhibiting marked
ellipticity.51 In ethene the value of L(r) at the charge concentration for the C-C interaction is 1.5 compared to 1.0 au in ethane,
while the two curvatures of L(r) at the critical point perpendicular to the C-C axis, which are equal in ethane, differ by a
factor of 2, the lesser one being directed perpendicular to the
molecular plane leading to an enhanced concentration of
electronic charge in the plane as evident in Figure 8.

The Electron Pair

Figure 8. Contour plots for L(r) and (r) for (a) ethane, (b) ethene in
the plane of the nuclei, (c) ethene in the plane perpendicular to (b) and
containing the C-C axis, and (d) ethyne. The central contour in (r)
in plot c of value 0.9 envelops two off-axis maxima of value 0.94 (Table
4). Plot d for (r) shows a cross section of the ring attractor.

The field for (r) is readily derived from that for L(r) in
ethene by replacing the bonded domain for the C-C interaction
in the atomic graph for L(r) by a linked pair of maxima in (r), thereby accounting for their characteristic sets differing by
one (3,-3) and one (3,-1). The remainder of the atomic graphs
is identical with a split pair of (3,-1) cps, as found in BH3,
linking the bonded maxima to carbon with those associated with
the protons. Of particular importance is the presence of two
(3,+1) cps situated on each side of the molecular plane above
and below each carbon atom. Their presence as holes in the
VSCC of carbon, along with their relative size, has been used
to account for the point and ease of nucleophilic attack in a
series of substituted ethenes.52
In ethyne a single symmetrically placed bonded charge
concentration links the two carbon atoms in L(r), while in (r)
this is replaced by a ring attractor,46 a connected set of (2,-2)
cps. Both fields have a bonded domain associated with each
proton, and the central attractor is linked to these by a ring of
connected (2,0) cps whose presence of also clear in Figure 8.
The charge concentrations associated with the C-C interactions

J. Phys. Chem., Vol. 100, No. 38, 1996 15409

Figure 9. Contour maps for L(r) and (r) for (a) N2, (b) CO, (c) H2CO in the plane of the nuclei, and (d) in the perpendicular plane
containing the C-O axis. Note the pronounced nonbonded domain
on C in CO evident in both fields in (b) and the very marked holes
or (3,+1) cps in the VSCC of the carbonyl carbon in (d).

in L(r) increase in the order ethane < ethene < ethyne, with
the electron localization domains of (r) exhibiting the reverse
order.
The isoelectronic molecules N2 and CO are also triply bonded
within the Lewis model, but neither exhibits a ring attractor as
found in ethyne. In N2 the two fields are homeomorphic, both
exhibiting an axial nonbonded domain on each atom that is
linked via a ring of (2,0) cps to a central axial bonded domain
(Figure 9). The fields are again homeomorphic in CO, with
the oxygen atom exhibiting the same structure as N in N2. The
carbon atom exhibits a very pronounced axial nonbonded
domain in both fields that is linked to the central bonded domain
by a ring of (2,0) cps.
In the H2CO molecule, (r) exhibits only a single axial
bonded domain between the C and O atoms, as opposed to the
two found in L(r). If the single domain in (r) is considered
shared by both atoms, the two fields are homeomorphic for the
C and O atoms. Only axial cps are found for the formally
doubly bonded CsO interaction. The atomic graph for carbon
in this molecule is homeomorphic with that for the L(r) field

15410 J. Phys. Chem., Vol. 100, No. 38, 1996

Figure 10. Contour maps of L(r) and (r) for (a) SO2, (b) SO3 in the
plane of the nuclei, and (c) SO3 in a perpendicular plane containing an
S-O axis. The central bonded contour of value 0.8 for (r) in (c)
encompasses two off-axis maxima of value 0.83 (Table 4).

for carbon in ethene. Most importantly, as noted previously,


L(r) for the carbonyl carbon exhibits very pronounced holes
or (3,+1) cps on each side of the molecular plane, which, as
for the C atoms in ethene, are the sites of nucleophilic attack.6
The holes are larger than in ethene (Figure 9), and their critical
points form an obtuse angle with the CsO bond axis and in
L(r) it equals 110, the angle found by Burgi and Dunitz53 to
be the most favored angle of approach of a nucleophile to a
carbonyl carbon as determined by crystallographic data. The
carbon atom of the CdO group in a carboxylic acid exhibits
the same atomic graphs as found in formaldehyde and is
characteristic of the carbonyl carbon.
The two fields are homeomorphic for the oxygen atom in
the carbonyl group (Table 4). It exhibits one bonded domain,
as it does in carbon monoxide, but the single axial nonbonded
domain found in CO is replaced by two pronounced in-plane
nonbonded domains. It is observed that when CO acts as a
terminal ligand in transition metal complexes, the oxygen VSCC
possesses a single axial nonbonded CC as in isolated CO, but
when acting as a bridging ligand, its nonbonded pattern is that
found in the carbonyl functional group.54
In a Lewis structure for SO2 that satisfies the octet rule for
both atoms, the average bond order linking them is 1.5, which
increases to 2.0 for a valence shell expanded to 10 on S. The
two fields are homeomorphic for S in this molecule and they
exhibit a single bonded maximum directed at each oxygen
(Table 4, Figure 10). This is shared with the oxygen in (r),
while in L(r) a separate bonded maximum is located in the
VSCC of each oxygen. Each bonded maximum of S is,
however, connected to the nonbonded one by (3,-1) cps

Bader et al.
symmetrically placed on each side of the plane of the nuclei in
the manner found for the carbon in the carbonyl group. The
double-bond character is also reflected in the significant
ellipticities of 1.4(S) and 3.1(O) for the bonded CCs, with the
value for the shared bonded maximum in (r) being 2.3. The
oxygen exhibits two nonbonded maxima in both fields.
In SO3 the corresponding Lewis structure predicts a bond
order of 1.33, increasing to 1.67 or 2.0 by allowing for an
expanded valence shell of 10 or 12 on S. The S-O link in
L(r) for this molecule has a single CC in each VSCC, as found
in SO2. (r), however, exhibits two bonded maxima symmetrically placed on each side of the plane of the nuclei, as
found for the C-C link in ethene, and because of this the two
fields are not homeomorphic. The bonded CCs in L(r) again
exhibit significant ellipticities of 2.2 on S and 2.5 on O (Figure
10), and the values of L(r) at the bonded maxima on S exceed
those found in SO2 (Table 4). The greater double-bond
character found in SO3 than in SO2 is reflected in the larger
ellipticity found for F(r) at the bond cp in the former: 0.261
compared to 0.155.
In summary, in some molecules such as ethene and SO3, a
double bond is represented in (r) in the form of two shared
bonded domains, while in others, such as the carbonyl group
and SO2, (r) exhibits a single bonded domain. Similarly, in
ethyne, the triple bond is represented by a ring of (2,-2)
attractors, while in N2 and CO only axial bonded and nonbonded
attractors are present. Double and triple bonds are represented
by axial cps in L(r), one on each atom in most cases, and by a
single shared maximum in others, as in ethyne and CO. A single
bond is, in general, represented by a single bonded domain in
(r), which appears to be shared by both atoms except when
bonded to hydrogen, in which case the bonded domain is
localized about the proton. In L(r), a single bond is represented
by a maximum in the VSCC of each atom in other than hydrides
and ionic or very polar species where a single maximum appears
in the VSCC of the electronegative atom. The cps in L(r)
exhibit radii characteristic of the corresponding valence shell
of the atom.
Molecules with Five Electron Pair Domains. The two
fields appear to be homeomorphic for the Cl atom in ClF3 with
the characteristic set [5,7,4]. The T-shaped geometry of this
molecule is a consequence of the two large nonbonded domains,
which, along with the equatorial bonded domain, are evident
in both fields in Figure 11a. By occupying the less crowded
equatorial positions, the angle formed between the two nonbonded cps can expand to a value well in excess of 120,
equaling 147 in L(r) and 144 in (r). The three bonded and
two nonbonded domains in L(r) exhibit the relative angular
positions and sizes anticipated on the basis of VSEPR, with
the axial ones being the smallest (Table 1). In spite of the fact
that both fields exhibit the number of edge and face cps requiring
five vertices, MORPHY is unable to locate the two axial bonded
domains in (r). The plots for both fields in the plane of the
three ligands are very similar in the regions of the Cl-F bonds
to the corresponding plots for ClF5 in Figure 2, with the roles
of the equatorial and axial ligands reversed. The absence of
the axial domains in (r) for ClF3 is a mathematical anomaly.
The remaining two examples, SOF4 and SF4 (Figure 11, Table
1), are molecules that, according to VSEPR theory, should
possess five electron pair domains as does ClF3, with the one
associated with the S-O linkage in SOF4 being large and
asymmetric since it represents two electron pairs. This is indeed
the case for L(r) in SOF4, but for (r) the domain for the S-O
linkage, as it is in SO3, is bifurcated in two symmetrically placed
about the C2 axis in the plane containing the equatorial ligands.

The Electron Pair

J. Phys. Chem., Vol. 100, No. 38, 1996 15411


5). These geometrical angles can be considered consequences
of the differences in the two related interdomain angles: b(O)S-b(e) ) 125 and b(O)-S-b(a) ) 98.3. There are secondary
differences between the two fields in the manner in which the
bonded domains are linked to one another.
The SF4 molecule has a single nonbonded domain in place
of the oxygen in SOF4. Both L(r) and (r) do indeed exhibit
one large nonbonded domain spread out in the equatorial plane
and four bonded domains. The nonbonded domain is linked to
the four bonded domains in both fields, and the fields would
be homeomorphic but for the appearance of an axial nonbonded
pseudomaximum in a region where L(r) is negative and the
density low in value. It does not appear as a maximum in the
diagrams for L(r), and it possesses one near-vanishing curvature.
It and the four (3,-1) cps linking it to the fluorines are readily
transformed by a bifurcation catastrophe into five corresponding
(3,-1) cps linking the fluorines and the two associated ring
cps found in the (r) field. The structure derived from these
five cps in L(r) is transformed into the structure defined by the
seven cps in (r) by the bifurcation of the central (3,-3) cp
into a (3,-1) and two ring cps. Thus, the [5,9,6] set of (r) is
derivable from the [6,8,4] set of L(r) by a bifurcation catastrophe
involving an unstable cp in a region of relatively low density.
Transition Metal Complexes. The Lewis model is not well
defined for heavy metal atoms: those from the third row onward
and the VSEPR model can fail.26 The topology of L(r) has
recently been used to determine the relative positions and sizes
of the electron domains in numerous examples of such
molecules, using large basis sets and electron correlation where
needed, as described in the original references.55,56 In all cases,
except for hydrides, q(M) > +2.0 and the outer ns electrons of
the metal atom M are transferred to the ligands, and one finds
the outer shell of the core of M containing the (n - 1)d electrons
to be significantly distorted. The distortions, which remarkably
are apparent in contours of the electron density with values as
high as 2.0 au,56 are made most evident in the terms of the
local CCs present in the Laplacian distribution of the outer shell
of the core of M, distortions that are already evident in the
alkaline earth dihydrides and dihalides beginning with Ca.
The pattern of electron localization determined in this manner
is indeed different from that observed for the main group
elements, in that the distortions are such as to produce electron
domains that are opposed to rather shared with or directed at
the ligands. This behavior is found even in molecules that
apparently follow the VSEPR rules, such as the trigonal
bipyramidal geometry found for VF5. While one cannot
distinguish between bonded or opposed CCs for the axial ligands
in this molecule, the three equatorial CCs are all ligand opposed.
VSEPR apparently fails for VH5 and V(CH3)5 in that calculations predict these molecules to possess square-based pyramidal
geometries,57 and one finds the electron domains to be larger

Figure 11. Contour maps of L(r) and (r) for ClF3 in the equatorial
plane containing the two nonbonded domains (a), for SOF4 in the plane
of the axial fluorines (b), for SOF4 in the plane of the equatorial
fluorines (c), and for SF4 in the plane of the equatorial fluorines (d).
The bonded contour shared with oxygen in (r) of plot c encompasses
two maxima of value 0.84 (Table 1). This maximum exhibits
considerable ellipticity in both fields, as does the corresponding
nonbonded domain on S in SF4 shown in (d), as described in the text.

In L(r) the corresponding single S-O domain is the largest of


the five bonded domains and exhibits a large ellipticity with
the charge concentration spread out toward the equatorial
ligands, a picture entirely consistent with (r) and one that
accounts for the angle formed by O with the axial ligands being
considerably smaller than that with the equatorial ligands (Table

TABLE 5: Bond Critical Point and Atomic Data for Molecules with Five and Six Electron Domainsa
molecule
ClF5, E ) -956.345 74, FeClFa ) 85.5
ClF3, E ) -757.626 88, FaClFe ) 86.9
SF6, E ) -994.229 80
SF4, E ) -795.314 32, FaSFa ) 171.0, FeSFe ) 102.2
SOF4, E ) -870.171 39, OSFa ) 97.2, OSFe ) 124.2

bond

RA-B

Fb

2Fb

Hb

q(A)

q(B)

l(A)

l(B)

Cl-Fa
Cl-Fe
Cl-Fa
Cl-Fe
S-F
S-Fa
S-Fe
S-O
S-Fa
S-Fe

2.962
3.076
3.173
2.955
2.905
3.087
2.871
2.621
2.971
2.855

0.2716
0.2264
0.1859
0.2519
0.2341
0.1821
0.2350
0.3635
0.2127
0.2427

-0.5154
-0.1303
+0.0652
-0.3802
+0.2333
-0.0155
+0.3458
+0.9570
+0.1552
+0.3998

-0.283
-0.185
-0.127
-0.274
-0.288
-0.182
-0.278
-0.550
-0.248
-0.292

+2.526
+2.526
+1.629
+1.629
+4.389
+2.914
+2.914
+4.313
+4.313
+4.313

-0.420
-0.527
-0.595
-0.440
-0.732
-0.722
-0.736
-1.395
-0.736
-0.723

0.86
0.86
0.91
0.91
0.88
0.90
0.90
0.88
0.88
0.88

0.93
0.94
0.95
0.93
0.94
0.94
0.95
0.92
0.95
0.94

Bond angles in degrees; all other quantities in atomic units except for l() which is dimensionless. The 6-311++G(2d,2p) basis was used for
geometry optimization and determination of the electron density using GAUSSIAN92.21 See Appendix for explanation of symbols.

15412 J. Phys. Chem., Vol. 100, No. 38, 1996

Bader et al.

TABLE 6: Bond Critical Point and Atomic Data for Hydridesa


AHn

energy

RA-H HAH

Fb

2Fb

Hb

q(A)

q(H)

l(A)

l(H)

LiH
BeH2
BH3
CH4
NH3

-7.986 16
-15.771 36
-26.399 18
-40.212 32
-56.218 58

0.0398
0.0978
0.1864
0.2894
0.3604

+0.156
+0.189
-0.258
-1.124
-1.889

-0.001
-0.050
-0.210
-0.326
-0.536

+0.912
+1.729
+2.112
+0.175
-1.047

-0.912
-0.865
-0.704
-0.044
+0.349

0.95
0.88
0.74
0.66
0.83

0.95
0.90
0.77
0.47
0.29

OH2

-76.057 43

3.307
2.515
2.243
2.044
1.886
107.9
1.776
106.3
1.696
3.622
3.225
2.981
2.785
2.656
95.4
2.506
94.1
2.390

0.3964

-3.106

-0.853

-1.254

+0.627

0.93

0.16

0.4043
0.0246
0.0545
0.0838
0.1220
0.1651

-4.006
+0.088
+0.220
+0.251
+0.179
-0.107

-1.091
+0.003
-0.006
-0.030
-0.082
-0.164

-0.779
+0.816
+1.614
+2.361
+2.895
+1.694

+0.779
-0.816
-0.807
-0.787
-0.724
-0.565

0.98
0.98
0.96
0.94
0.91
0.91

0.10
0.90
0.89
0.84
0.77
0.65

0.2210

-0.657

-0.228

+0.284

-0.142

0.93

0.45

0.2571

-0.787

0.254

-0.222

+0.222

0.97

0.33

FH
NaH
MgH2
AlH3
SiH4
PH3

-100.056 47
-162.380 56
-200.731 93
-243.640 96
-291.257 25
-342.481 92

SH2

-398.706 38

HCl

-460.098 80

See footnote a to Table 5.

TABLE 7: Bond Critical Point and Atomic Data for Fluorides and PCl3a
AFn

energy

RA-H HAH

Fb

2Fb

Hb

q(A)

q(H)

l(A)

l(H)

LiF
BeF2
BF3
CF4
NF3

-106.978 94
-213.760 38
-323.308 00
-435.787 77
-352.661 26

0.0763
0.1498
0.2917
0.3186
0.3773

+0.768
+1.480
+0.443
-0.153
-0.859

+0.024
-0.014
-0.241
-0.528
-0.453

+0.940
+1.807
+2.576
+2.964
+1.086

-0.940
-0.904
-0.859
-0.741
-0.362

0.96
0.90
0.83
0.70
0.74

0.99
0.99
0.97
0.95
0.93

OF2

-273.550 44

0.3697

-0.079

-0.318

+0.332

-0.166

0.84

0.93

F2
NaF
MgF2
AlF3
SiF4
PF3

-198.747 55
-261.356 86
-398.691 03
-540.580 79
-687.119 44
-639.264 51

0.3669
0.0518
0.0834
0.1215
0.1628
0.1793

+0.154
+0.453
+0.859
+1.210
+1.351
+0.936

-0.306
+0.015
+0.021
+0.003
-0.054
-0.122

0.000
+0.941
+1.826
+2.647
+3.428
+2.510

0.000
-0.941
-0.913
-0.882
-0.857
-0.837

0.93
0.99
0.98
0.96
0.94
0.93

0.93
0.99
0.99
0.98
0.97
0.96

SF2

-596.412 71

2.953
2.574
2.444
2.447
2.496
102.9
2.526
103.5
2.512
3.647
3.279
3.050
2.912
2.922
97.3
2.970
97.2
2.882
100.2
3.889
100.4

0.1952

+0.344

-0.201

+1.427

-0.714

0.94

0.95

0.2768

-0.496

-0.344

+1.764

-0.382

0.92

0.93

0.1270

-0.164

-0.106

+1.776

-0.592

0.90

0.97

ClF2

-657.853 56

PCl3

-1719.342 60

See footnote a to Table 5.

TABLE 8: Bond Critical Point and Atomic Data for Single, Double, and Triple Bondsa
molecule
C2H6, E ) -79.257 33, CCH ) 111.2
C2H4, E ) -78.061 42, CCH ) 121.6
C2H2, E ) -76.846 95
N2, E ) -108.981 24
CO, E ) -112.776 38
H2CO, E ) -113.908 29, HCO ) 121.8
(CH3)2O, E ) -154.120 20, HCO ) 107.8,b HCO ) 111.4c
SO2, E ) -547.271 32, OSO ) 118.8
SO3, E ) -622.118 46
a

bond

RA-B

Fb

2Fb

Hb

q(A)

q(B)

l(A)

l(B)

C-C
C-H
C-C
C-H
C-C
C-H
N-N
C-O
C-O
C-H
C-O
C-Hb
C-Hc
S-O
S-O

2.881
2.048
2.485
2.029
2.229
1.992
2.015
2.085
2.227
2.063
2.628
2.040
2.054
2.644
2.625

0.2496
0.2908
0.3710
0.3000
0.4407
0.3080
0.7649
0.5332
0.4475
0.3003
0.2730
0.3029
0.2964
0.3343
0.3492

-0.614
-1.132
-1.342
-1.216
-1.560
-1.316
-3.766
+0.739
+0.139
-1.227
-0.343
-1.232
-1.178
+1.205
+1.041

-0.205
-0.329
-0.475
-0.343
-0.709
-0.360
-1.600
-1.041
-0.833
-0.335
-0.426
-0.345
-0.333
-0.461
-0.507

+0.183
+0.183
+0.035
+0.035
-0.136
-0.136
0.000
+1.357
+1.292
+1.292
+0.776
+0.776
+0.776
+2.712
+4.105

+0.183
-0.061
+0.035
-0.017
-0.136
+0.136
0.000
-1.357
-1.271
-0.010
-1.288
-0.022
-0.054
-1.356
-1.368

0.65
0.65
0.67
0.67
0.68
0.68
0.78
0.89
0.66
0.66
0.65
0.65
0.65
0.90
0.87

0.65
0.47
0.67
0.45
0.68
0.39
0.78
0.92
0.91
0.47
0.89
0.47
0.48
0.91
0.92

See footnote a to Table 5. b H in plane of C-O-C. c H out of plane.

and more diffuse in these cases than those found in VF5.56 The
ligand-opposed nature of the CCs in the outer core, as defined
by L(r) for the V atom, is displayed in Figure 12 for VH5 where
q(V) ) +1.7e. Also shown is a corresponding display of (r)
for this molecule, and the atomic graphs for V are clearly
homeomorphic in terms of the electron domains that they define.

The same homeomorphism is evident in corresponding displays


for CrOF4, q(Cr) ) +2.9e, with the large ligand-opposed
electron domain generated by the presence of the oxygen atom
being evident in both fields. Molecules with nonbonded CCs
behave similarly, as illustrated in the previously given display55
of the four tetrahedrally arranged CCs present in the outer core

The Electron Pair

Figure 12. Envelope maps for the minimum-energy geometries: VH5


L(r) (a) and (r) (b); CrOF4 L(r) (c) and (r) (d). The function values
are 0.15 au in (a), 0.85 in (b), 21 au in (c), and 0.83 in (d). The
corresponding geometries for the same orientation of the nuclei
indicated in the insets make clear the ligand-opposed nature of the
electron domains in the outer shell of the cores of the transition metal
atoms. Note that the domain induced by and opposed to the oxygen
in CrOF4 is larger in both fields than are those induced by the fluorines.

of Ba in BaH2 with (r) exhibiting a similar pattern of two


nonbonded and two ligand-opposed CCs. The closed-shell core
of Zn in ZnCl42-, on the other hand, is more nearly spherical,
and the four maxima that are present in L(r) are interior to the
outer limits of the Zn atomic surface, are of limited spatial
extent, and are tetrahedrally disposed in the manner of bonded
CCs found in main group elements.58
This parallel behavior of the two fields for the transition metal
atoms is general and firmly establishes that the Lewis model,
as implemented in terms of spatially localized electron pair
domains, is different for atoms with vacancies in the (n - 1)d
shell from that found for main group elements.
Summary. An overview of the figures comparing the L(r)
and (r) fields convinces one of the essential similarity in their
structures. A wide spectrum of atomic interactions is represented, and, in nearly all cases, one finds a homemorphism in
the number of electron pair domains that they define for a central
atom interacting with a set of ligands. In general, the atomic
graph of L(r) is more characteristic of a given atom than is that
for (r). In ionic and very polar interactions, both fields in
general exhibit a single bonded domain on each ligand, this
behavior being obtained for the (r) field in all hydrides. In
shared interactions, a single bonded domain of (r) is, in general,
replaced by a pair of domains in L(r), one in each VSCC of the
bonded atoms. Multiple bonds in L(r) are represented by single
domains, the multiplicity of bonding being reflected in their
magnitudes and ellipticities, similar to the topology exhibited
by F(r). A double bond in (r) can be represented by two
domains and a triple bond in a linear system by a ring attractor,
but this behavior is not general. The homeomorphism exhibited
by the two fields for transition metal molecules demonstrates
the presence of a new pattern for the spatial pairing of electrons
in the form of ligand-opposed domains in the outer shell of the
metal atom core.
By accepting the presence of the doubling of bonded domains
in some instances and the replacement of a shared domain by

J. Phys. Chem., Vol. 100, No. 38, 1996 15413


a ring attractor in another, the two fields offer a consistent view
of the regions in a molecule that are dominated by the presence
of electron pairing and their angular distribution in the vicinity
of a single atom, as envisaged in the VSEPR model or as found
for transition metal atoms. In general, the radial distances at
which the localization domains are found are larger in (r) than
in L(r). The cps in L(r) are determined entirely by the properties
of the electron distribution. The agreement between L(r) and
(r) in their predicted angular distribution of the localization
domains would suggest that the element of arbitrariness
introduced into the definition of (r)9 by measuring the
smallness of relative to that for a uniform electron gas, while
affecting the radial distances, does not affect the number or
angular orientation of the domains.
From the examples given previously and here, it appears that
L(r) is superior to (r) in predicting the location and magnitude
of the lumps and holes created in real space by the partial
pairing of electrons, which underlies the Lewis model of acidbase reactivity. The Laplacian of the electron density is the
only property of F(r) directly related to an energy density. It
provides, through eq 4, a measure of the local potential and
kinetic energy densities of the electrons, thereby yielding the
physical basis underlying an acid-base reaction: a base site is
a spatial region exhibiting an excess of potential energy in which
the extent of electron pairing is a maximum whereas an acid is
one exhibiting an excess of kinetic energy for which electron
pairing is a minimum in the quantum shells of interest.
The single electron domain shared by two bonded atoms that
is exhibited by (r), on the other hand, is more useful for
comparing the bonding through a series of similar compounds
a series of substituted X-O-X molecules, for example. That
the homeomorphism is not complete means that the two fields
complement one another. L(r), in general, reflects the VSEPR
model of associating a set of electron domains with a central
atom, while (r) generally exhibits a single shared domain
between two bonded atoms. Thus, L(r) associates a bonded
CC in the VSCC of each carbon in a C-C bond, while (r)
yields a single shared domain (Figure 8). Clearly, one cannot
in such cases associate a pair of electrons with every CC in
L(r), but must instead do so for the linked pair of CCs. What
is represented in (r) as a single bonded domain appears as a
bifurcated set of CCs in L(r) linked by a ridge of charge
concentration (Figure 8). There are counterexamples. In ClF3
the equatorial fluorines in (r) exhibit a bonded maximum in
addition to the one found on Cl, while L(r) exhibits only the
CC within the VSCC of Cl. Neither field yields an electron
count in terms of the number of local maxima it exhibits. The
description provided by L(r) is, however, nonarbitrary and
indicates that the pairing phenomenon associated with bonding
extends well into the valence shell of each atom, as indicated
by the radial values listed in tables.
Conclusions
The pairing of electrons is a consequence of the spatial
localization of an electron of a given spin, as determined by a
corresponding localization of its Fermi hole.4 While the
quantitative measures of electron localization made possible by
this understanding show that the pairing of electrons is less
pronounced than envisaged in the original Lewis model, there
is evidence of partial condensation of the pair density to yield
spatial domains with pair populations approaching unity.4,8
Further evidence is provided by the success of the VSEPR model
in predicting the geometry of main group elements by using
arguments of electron pairing based upon the operation of the
exclusion principle.26 The recovery of this model in terms of

15414 J. Phys. Chem., Vol. 100, No. 38, 1996

Bader et al.

the topology of the Laplacian of the electron density is


symbiotic: it provides a physical basis for the VSEPR model
and, because of the success of this model, it establishes an
empirical correspondence between the localized charge concentrations defined by L(r) with the electron domains assumed
in the VSEPR model. Thus, it establishes a connection between
L(r) and the pair density.
The measure of electron localization provided by Becke and
Edgecombes9 ELF, the function (r), is directly related to the
pair density. The function appearing in (r), the difference
between the many- and one-electron kinetic energy densities,
provides a measure of the conditional pair probability for an
electron of given spin. Electron localization is related to the
smallness of ,9 a condition which, as previously demonstrated,42 is once again a direct result of the localization of the
Fermi hole.
Thus, the general homeomorphism between the (r) and L(r)
fields demonstrated here establishes a physical link between the
protrayal of electron localization in terms of the charge
concentrations of L(r) and the associated properties of the pair
density. By extension, it provides justification for the qualitative
arguments used to rationalize the VSEPR model of molecular
geometry.
Acknowledgment. We thank Professor R. J. Gillespie for
useful discussions concerning this paper.
Appendix
Definition of the Fermi Hole Density. The motions of the
same-spin electrons are correlated, and the pair density written
for the R-spin electrons obtained from a single-determinant wave
function is

FRR(r1,r2) ) 1/2FR(r1){FR(r2) + hR(r1,r2)}

(A1)

where hR(r1,r2) is the density of the Fermi hole for an R-spin


electron. It is the negative of Slaters exchange charge density60
and expressed in terms of the spin orbitals i is
R

R
hR(r1,r2) ) -{*
i(r1)i(r2)*
j(r2)j(r1)}/F (r1) (A2)

Unlike the pair density for electrons of opposite spin in eq 1,


the same-spin pair density in eq A1 is decreased from the simple
product of single particle densities in a manner described by
the term in curly brackets. This term, R(r2) in eq 2 of the text
and displayed in Figure 1, is usually interpreted as a conditional
probability.3,9 It has two important properties:
(a) It equals zero for r1 ) r2 because hR(r1,r1) equals -FR(r1), corresponding to the complete removal of same-spin density
from the position of the reference electron. The density of the
Fermi hole may approach -FR(r1) for values of r1 * r2, and
the conditional probability will then approach zero over the
region of space where this near-equality is obtained, leading to
a correspondingly near-exclusion of all other same-spin electrons, as illustrated in Figure 1a,c.
(b) The integral of the conditional probability over r2 equals
NR - 1, the integration of the Fermi hole density resulting in
the removal of one R-spin electron, reducing their total number
NR by 1. In this manner, the product of single particle densities
is corrected for the self-pairing (and self-repulsion) of the
electrons to yield 1/2NR(NR - 1) pairs.
A bound electron cannot be localized, and r1 and r2 cannot
be strictly interpreted as electronic coordinates. Interpretation
of the vanishing of the conditional probability for r1 ) r2

requires that the charge of the reference electron at r1 be equal


to FR(r1) dr1 rather than to (r - r1) dr1, as it would be if indeed
the whole electronic charge was localized at r1. Thus, r1 and
r2 denote the coordinates of two points in a six-dimensional
space, and the term in curly brackets describes the manner in
which the remaining charge of an electron with density FR(r1)
at r1 is spread out in the space described by r2, decreasing the
same-spin density at each point in this space by an amount
determined by the Fermi hole. The Fermi correlation is so
demanding and so pronounced that it is well described at the
single-determinant approximation to the wave function in
Hartree-Fock theory, the level of theory used here. All of the
ideas, however, apply to the exact wave function as well.
Tables of Atomic and Bond Critical Point Data. Tables
5-8 compile the energy, geometrical parameters, bond critical
point data, atomic populations q(A), and electron localizations
l(A). Quantities subscripted with a b refer to their value at the
bond critical point. Hb is the energy density at the bond critical
point61 and is equal to the sum of the kinetic and potential energy
densities, G(r) and (r), appearing in the expression for the local
virial theorem (eq 4). According to this same theorem, Hb(r)
is equal to -K(r), the alternative expression for the kinetic
energy density.8 Unlike the sign of 2Fb, which is determined
by the local virial expression in eq 4 wherein the potential energy
is compared with twice the kinetic contribution, Hb is determined
by the energy density itself and is thus negative for all
interactions that result from the accumulation of density at the
bond cp.
The bond path has been given added physical significance
with the demonstration of a structural homeomorphism between
the topology of F(r) and that of the virial field (r), the potential
energy density appearing in the local virial theorem eq 4. This
homeomorphism means that a molecular graph defining a
molecualr structure is mirrored by a corresponding virial graph
and the lines of maximum density linking bonded nuclei in F(r)
are matched by a set of lines of maximally negative potential
energy density in (r).62
References and Notes
(1) Lewis, G. N. J. Am. Chem. Soc. 1916, 38, 762.
(2) If one could increase the spin angular momentum quantum number
for an electron from S ) 1/2 to S ) 3/2, or in a universe where this was so,
there would be four observable spin components. One could construct an
antisymmetrized orbital wave function with up to four electrons occupying
each space orbital, allowing for the possibility of observing the spatial
quadrupling of electrons. The rows of the periodic table would be
correspondingly longer and Be would be the first of the inert gases.
(3) McWeeny, R. ReV. Mod. Phys. 1960, 32, 335. McWeeny, R.;
Sutcliffe, B. T. Methods of Molecular Quantum Mechanics; Academic
Press: London, 1969; pp 100.
(4) Bader, R. F. W.; Stephens, M. E. J. Am. Chem. Soc. 1975, 97,
7391.
(5) Bader, R. F. W.; Streitwieser, A.; Neuhaus, A.; Laidig, K. E.;
Speers, P. J. Am. Chem. Soc. 1996, 118, 4959.
(6) Bader, R. F. W.; MacDougall, P. J.; Lau, C. D. H. J. Am. Chem.
Soc. 1984, 106, 1594.
(7) Bader, R. F. W.; Gillespie, R. J.; MacDougall, P. J. J. Am. Chem.
Soc. 1988, 110, 7329.
(8) Bader, R. F. W. Atoms in MoleculessA Quantum Theory; Oxford
University Press: Oxford, UK, 1990.
(9) Becke, A. D.; Edgecombe, K. E. J. Chem. Phys. 1990, 92, 5397.
(10) Becke, A. D. Int. J. Quant. Chem. 1983, 23, 1915.
(11) Maslen, C. W. Proc. Phys. Soc. (London) 1956, A69, 734.
(12) Luken, W. L. Croat. Chim. Acta 1984, 57, 1283. Luken, W. L.;
Beratan, D. N. Theor. Chim. Acta 1982, 61, 265. Luken, W. L.; Culberson,
J. C. Int. J. Chem.: Symp. 1982, 16, 265.
(13) Daudel, R. C. R. Acad. Sci. (Paris) 1953, 237, 601.
(14) Daudel, R.; Brion, H.; Odiot, S. J. Chem. Phys. 1955, 23, 2080.
Odiot, S. Cahiers Phys. 1957, 81, 1.
(15) Aslangul, C.; Constanciel, R.; Daudel, R.; Kottis, P. AdV. Quant.
Chem. 1972, 6, 93.
(16) Aslangul, C. C. R. Acad. Sci. (Paris) 1971, 272B, 1.

The Electron Pair


(17) Daudel, R.; Bader, R. F. W.; Stephens, M. E.; Borrett, D. S. Can.
J. Chem. 1974, 52, 1310.
(18) Bader, R. F. W.; Stephens, M. E. Chem. Phys. Lett. 1974, 26, 445.
(19) (a) Bader, R. F. W. Phys. ReV. 1994, B49, 13348. (b) Bader, R. F.
W.; Popelier, P. L. A.; Keith, T. A. Angew. Chem., Int. Ed. Engl. 1994, 33,
620.
(20) Biegler-Konig, F. W.; Bader, R. F. W.; Tang, T.-H. J. Comput.
Chem. 1982, 3, 317. The most recent version V is available from R. F. W.
Bader.
(21) Frisch, M. J.; Trucks, G. W.; Head-Gordon, M.; Gill, P. M. W.;
Wong, M. W.; Foresman, J. B.; Johnson, B. G.; Schlegel, H. B.; Robb, M.
A.; Replogle, E. S.; Gomperts, R.; Andres, J. L.; Raghavachari, K.; Binkley,
J. S.; Gonzalez, C.; Martin, R. L.; Fox, D. J.; DeFrees, D. J.; Baker, J.;
Stewart, J. J. P.; Pople, J. A. GAUSSIAN92; Gaussian, Inc.: Pittsburgh,
PA, 1992.
(22) Ebbing, D. D.; Henderson, R. C. J. Chem. Phys. 1965, 42, 2225.
Davidson, E. R. ReV. Mod. Phys. 1972, 44, 451. Ruedenberg, K.; Mehler,
E. L. J. Chem. Phys. 1970, 52, 1206. Kutzelnigg, W. Fortshr. Chem.
Forsch. 1973, 41, 31. Sinanoglu, O.; Skutnik, B. Chem. Phys. Lett. 1968,
1, 699.
(23) Daudel, R.; Stephens, M. E.; Kapuy, E.; Kozmutza, C. Chem. Phys.
Lett. 1976, 40, 194.
(24) Bader, R. F. W.; Nguyen-Dang, T. T.; Tal, Y. Rep. Prog. Phys.
1981, 44, 893.
(25) Bader, R. F. W.; Gillespie, R. J.; MacDougall, P. J. In Molecular
Structure and Energetics; Liebman, J. F., Greenberg, A., Eds.; VCH
Publishers Inc.: New York, 1989; Vol. 11.
(26) Gillespie, R. J.; Nyholm, R. S. Q. ReV. Chem. Soc. 1957, 11, 239.
Gillespie, R. J. Molecular Geometry; Van Nostrand Reinhold: London,
1972. Gillespie, R. J. Chem. Educ. 1963, 40, 295. Gillespie, R. J.; Hargittai,
I. The VSEPR Model of Molecular Geometry; Allyn and Bacon and Prentice
Hall International: Boston, MA, and London, respectively, 1991.
(27) Bader, R. F. W.; Essen, H. J. Chem. Phys. 1984, 80, 1943.
(28) Sagar, R. P.; Ku, A. C.; Smith, V. H.; Sinas, A. M. J. Chem. Phys.
1988, 88, 4367.
(29) Shi, Z.; Boyd, R. J. J. Chem. Phys. 1988, 88, 4375.
(30) Collard, K.; Hall, G. G. Int. J. Quant. Chem. 1977, 12, 623.
(31) Palis, J.; Smale, S. Pure Math. 1970, 14, 223.
(32) Thom, R. Structural Stability and Morphogenesis; W. A. Benjamin: Reading, MA, 1975.
(33) Aray, Y.; Bader, R. F. W. Surf. Sci. 1996, 351, 233.
(34) Tsirelson, V.; Zou, P. F.; Tang, T.-H.; Bader, R. F. W. Acta
Crystallogr. 1995, A51, 143.
(35) Aray, Y.; Rosillo, F.; Murgich, J. J. Am. Chem. Soc. 1994, 116,
10639. Aray, Y.; Rodriguez, J.; Murgich, J.; Ruette, F. J. Phys. Chem.
1993, 97, 8393.
(36) Tang, T.-H.; Hu, W.-J.; Yan, D.-Y.; Cui, Y.-P. J. Mol. Struct.
(Theochem) 1990, 207, 327.
(37) Platts, J. A.; Howard, S. T.; Wozniak, K. J. Org. Chem. 1994, 59,
4647.

J. Phys. Chem., Vol. 100, No. 38, 1996 15415


(38) Alcam, M.; Mo, O.; Yanez, M.; Abboud, J.-L. M. J. Phys. Org.
Chem. 1991, 4, 177.
(39) Howard, S. T.; Platts, J. A. J. Phys. Chem. 1995, 99, 9027.
(40) Slee, T.; Bader, R. F. W. J. Mol. Struct. (Theochem) 1992, 87,
173.
(41) Abboud, J.-L. M.; Canada, T.; Homan, H.; Notario, R.; Cativiela,
C.; Diaz de Villegas, D.; Bordede, M. C.; Mo, O.; Yanez, M. J. Am. Chem.
Soc. 1992, 114, 4728.
(42) Tal, Y.; Bader, R. F. W. Int. J. Quant. Chem., Quant. Chem. Symp.
1978, 12, 153.
(43) (a) Bader, R. F. W.; Preston, H. J. T. Int. J. Quant. Chem. 1969, 3,
327. (b) von Weiszacker, C. F. Z. Phys. 1935, 96, 431.
(44) Anderson, M. H.; Ensher, J. R.; Matthew, M. R.; Wieman, C. E.;
Cornell, E. A. Science 1995, 269, 198.
(45) (a) Savin, A.; Jepsen, O.; Flad, J.; Andersen, O. K.; Preuss, H.;
von Schnering, H. G. Angew. Chem., Int. Ed. Engl. 1992, 31, 187. (b)
Savin, A.; Becke, A. D.; Flad, J.; Nesper, R.; Preuss, H.; von Schnering,
H. G. Angew. Chem., Int. Ed. Engl. 1991, 30, 409.
(46) Silvi, B.; Savin, A. Nature 1994, 371, 683.
(47) Mezey, P. In ReViews in Computational Chemistry; Lipkowitz, K.
B., Boyd, D. B., Eds.; VCH: New York, 1990; Vol. 1.
(48) Fassler, T. F.; Hausserman, U.; Nesper, R. Chem. Eur. J. 1995, 1,
625.
(49) Morphy96 with manual by P. L. A. Popelier is available from
QCPE; MORPHY1.0 is available from CPC program library (Queens
University, Belfast: cpc@qub.ac.uk) or from the author (pla@umist.ac.uk).
A manual is published in Comput. Phys. Commun. 1996, 93, 212. The
integration technique is described in Mol. Phys. 1996, 87, 1169.
(50) Popelier, P. L. A. Chem. Phys. Lett. 1994, 228, 160.
(51) Gillespie, R. J.; Bytheway, I.; DeWitte, R. S.; Bader, R. F. W. Inorg.
Chem. 1994, 33, 2115.
(52) Carroll, M. T.; Cheeseman, J. R.; Osman, R.; Weinstein, H. J. Phys.
Chem. 1989, 93, 5120.
(53) Burgi, H. B.; Dunitz, J. D. Acc. Chem. Res. 1983, 16, 153.
(54) Bo, C.; Sarasa, J. P.; Poblet, J. M. J. Phys. Chem. 1993, 97, 6362.
(55) Bytheway, I.; Gillespie, R. J.; Tang, T.-H.; Bader, R. F. W. Inorg.
Chem. 1995, 34, 2407.
(56) Gillespie, R. J.; Bytheway, I.; Tang, T.-H.; Bader, R. F. W. Inorg.
Chem. 1996, 35, 3954.
(57) Kang, S. K.; Tang, H.; Albright, T. A. J. Am. Chem. Soc. 1993,
115, 1971.
(58) Bader, R. F. W.; Gamow, G.; Tang, T.-H. Unpublished work.
(59) Gillespie, R. J.; Johnson, S. Inorg. Chem., submitted.
(60) Slater, J. C. Phys. ReV. 1951, 81, 385.
(61) Cremer, D.; Kraka, E. Croat. Chim. Acta 1984, 57, 1259. Cremer,
D.; Kraka, E. Angew. Chem., Int. Ed. Engl. 1984, 23, 627.
(62) Keith, T. A.; Bader, R. F. W.; Aray, Y. Int. J. Quantum Chem.
1996, 57, 183.

JP961297J

Você também pode gostar